Leadership in AI? Watch Moneyball

We are sitting at LA LIVE in downtown Los Angeles with a few friends going to the Consumer Electronics Show (CES) in Las Vegas this coming week. The LA Kings are playing at the Staples Center, and we talk around a beer about the emergence of Artificial Intelligence (AI) in traditional sectors. A cohort of 35 start-ups from France will be showcased at the show. Despite their strong technical background, they have an uphill battle because AI transforms businesses and it is as much about people as technology.

The adoption of data science in non-tech sectors started in the US in professional sports over 15 years ago. Do you remember the movie Moneyball? The movie starring Bratt Pitt in the role of Billy Bean who integrated data mining into the DNA of at the Oakland A’s in 2002 after losing his 3 star players. In a Forbes article, his experience shows why business leaders can’t leave data science to the data scientists no matter how talented they are. The plot of the film summarizes the 4 phases of transformation:

1. Develop a clear understanding of the problem with the leadership

2. Find a data model and collect the relevant data within the organization

3. Sell the solution to the operational team in a way that is compatible with their day-to-day constraints

4. Monitor the relevance of the model and make adjustments if needed

Brad Pitt plays Billy Bean who understands that he cannot compete with big teams: “ The problem we’re trying to solve is that there are rich teams and there are poor teams. Then there’s 50 feet of crap, and then there’s us.”

Defining the right problem is difficult. It needs to boil down to a simple question that is clearly stated and can lead to a measurable objective. This means that the leadership of an organization must be aware of its situation, and that the pain is felt enough to want to take action.

In MoneyBall, Oakland A’s are confronted with a new reality after losing closely to the New York Yankees in the 2001 playoffs. In the off-season they lose their 3 star players. It hurts. The General Manager Billy Bean asks for more money to compete and the Owner tells him to manage a team the A’s can afford next season.

Billy Bean starts the process to define the question by explaining to the scouts that they cannot think as usual. They will not find another Jason Giambi. Billy Bean doesn’t need to understand the math, but bears the responsibility to frame the problem correctly. The question becomes: “How do we win ball games next season with a third of the budget of the Yankees?”

Jonah Hill plays the role of a recent Ivy League graduate who explains to Billy Bean how data can help find undervalued players due to a number of biases in the world of baseball: “ I believe that there is a championship team of twenty-five people that we can afford, because everyone else in baseball undervalues them.”

Billy Beane goes on a trip and negotiates with other organizations. He notices a low level staffer that explains that winning is not so much about players but about runs. And to get runs, a team needs to get on base.

The character played by Jonah Hill explains to Billy Beane that many players are undervalued because of biases: “ People who run ball clubs, they think in terms of buying players. Your goal shouldn’t be to buy players, your goal should be to buy wins. And in order to buy wins, you need to buy runs.” He becomes the GM’s first hire because he uses a proven data model called “Sabermetrics” to identify players that can make a good team at a low cost.

The difficulty for most organizations is to have access to good data (quality) and enough of it (quantity). The quality of the data depends on the relevance to the business question. For instance, sampling requirements and the list of useful of metrics will depend on the business question. That can differ from existing operational practices. There can be a cost to collect more data (“digital transformation”) within an organization. Or privacy issues can limit access to enough data to run a model.

Luckily for the GM of Oakland A’s, collegiate and professional sports kept detailed statistics on each player. They are widely available as players are acquired during drafts and traded frequently within the League. Sabermetrics can then predict runs for various team configurations, and create a list of players who can be acquired at low cost and still make a team winning games.

Billy Beane talking to the payers including David Justice and Scott Hatteberg. David: “How you likin’ first base, man?” Scott: It’s coming along. Picking it up. You know, tough transition, but I’m starting to feel better with it.

After a data model is chosen, tested and implemented, comes the most important phase: make an impact. Having a good data model is not enough if the operational team is not using it. In baseball that is the players and the coach. In MoneyBall, the coach rejects the new model by fear of negatively impacting his career and becoming a joke in the profession.

That is when the GM and the Data Scientist decide to talk to the players and travel on the road with them. They even enlist the service of the veteran player David Justice to “get on base” and implement the new plan.

Introducing a data model in an organization requires thinking first about people. Processes will need to be adapted to their day-to-day constraints so they see value in them. For David Justice, it is about having a meaningful career end and solidifying his legacy as Hall-of-Famer. He buys into the plan.

The film takes the time to show how the baseball team slowly jells. The leaders are committed to support the players in using the new approach and taking into account the data fed back to them. In general, staff needs to be trained to use new insights. They need to understand the limits of the model, and how to include the information in their day-to-day decisions.

Billy Beane: “Jeremy, you’ve been traded to the Phillies. This is Ed Wade’s number. He’s the GM. He’s expecting your call. Buddy will help you with the plane flight. You’re a good ballplayer, and we wish you the best.”

Over time, once the model is being actually used, changes may appear. It could be bad data, a change in factory line requiring a new metric, or simply an HR problem. Key Performance Indicators (KPI’s) should be built into the model to monitor its relevance and whether the organization is in track to achieve its goal.

Should the KPI’s diverge, the leadership should react by understanding the reasons and launching corrective actions. In MoneyBall, after the model is put in use and seems to start working, one player is not statistically behaving as predicted. After looking into the issue, Billy Bean discovers that this player is bad-tempered and creates a wrong dynamic in the team.

This was not part of the training model, and thus not predicted. In MoneyBall, Billy Bean decides to trade him and get the coach in line. The team then starts to go on a streak of 21 games won in a row. History. The A’s make the playoffs.

Sabermetrics changed the sports of professional sports for ever. MIT Sloan is hosting the 13th edition of the Sports Analytics Conference in Boston on March 1, 2019. The Boston Red Sox adopted the same data driven approach and beat the Curse of the Bambino. They won the World Series in 2004.